Fix tests that were setting safebrowsing factories without unsetting them.

chrome/browser/safe_browsing/safe_browsing_blocking_page_unittest.cc

Index: chrome/browser/safe_browsing/safe_browsing_blocking_page_unittest.cc
diff --git a/chrome/browser/safe_browsing/safe_browsing_blocking_page_unittest.cc b/chrome/browser/safe_browsing/safe_browsing_blocking_page_unittest.cc
index 5f4d322057be7dea857781186165f3b0403ff5f9..29d13f08d4dd698a1fae4f652c96000bc0ee847a 100644
--- a/chrome/browser/safe_browsing/safe_browsing_blocking_page_unittest.cc
+++ b/chrome/browser/safe_browsing/safe_browsing_blocking_page_unittest.cc
@@ -126,16 +126,16 @@ class SafeBrowsingBlockingPageTest : public ChromeRenderViewHostTestHarness {
     ui_manager_ = new TestSafeBrowsingUIManager(NULL);
   }
 
-  virtual void SetUp() {
+  virtual void SetUp() OVERRIDE {
     ChromeRenderViewHostTestHarness::SetUp();
     SafeBrowsingBlockingPage::RegisterFactory(&factory_);
-    MalwareDetails::RegisterFactory(NULL);  // Create it fresh each time.

[Scott Hess - ex-Googler, 2013/10/31 22:07:20]

Shouldn't this move to TearDown() rather than being deleted?  Or is it just not
used here?  In that case can malware_details.h be removed?

[mattm, 2013/10/31 22:39:23]

I believe it is just not being used. (Nothing was setting it.)

Well looked up the history, I guess it was part of
https://code.google.com/p/chromium/issues/detail?id=110594 ...
I moved it to TearDown and added a better comment, but that approach of having
every test try to clean up after any singletons it happens to activate seems
kind of doomed though.  

     ResetUserResponse();
   }
 
-  virtual void TearDown() {
+  virtual void TearDown() OVERRIDE {
     // Release the UI manager before the BrowserThreads are destroyed.
     ui_manager_ = NULL;
+    SafeBrowsingBlockingPage::RegisterFactory(NULL);
     ChromeRenderViewHostTestHarness::TearDown();
   }